Readiness: Monitoring That Shortens the Ransomware Window
The rest of this pillar has been about signals. It covers what a ransomware intrusion looks like at each stage, from the first quiet check-in to the last loud minute of encryption. Signals are only half of readiness. The other half is the boring, organizational part. It covers which of those signals you actually have alerts for, and who those alerts reach at three in the morning. It covers whether your backups have been proven restorable rather than assumed. And it covers whether the two people who will handle the first hour have ever practiced it. Most small teams have read about the signals. Far fewer have turned them into a program.
This guide is the program. It lays out the detection set as a single list of eight signals. Each signal comes with the instrument that provides it and the alert that should exist for it. The guide treats backup verification as a monitored control rather than a good intention. It argues that this is the one category of alert worth waking up for, and shows how to route it. And it describes the tabletop drill that turns a checklist into muscle memory. It ends with a readiness checklist you can score yourself against honestly. The goal is to shrink the window between foothold and detection from days to minutes — and to make recovery, if it is ever needed, boring.
In this guide
- The goal: days to minutes, and boring recovery
- The eight-signal detection set
- Backup verification as a monitored control
- Around-the-clock routing for the top signals
- The first-hour tabletop drill
- The readiness checklist
- Where to go next
The goal: days to minutes, and boring recovery
Readiness has two measurable outcomes, and it helps to name them before building anything. The first is time to detection: how long an intrusion runs on your network before a human knows about it. Without a program, that is however long the quiet stages last — usually days — and detection happens at the ransom note. With a program, it is however long the first monitored signal takes to fire and reach someone, which should be minutes. Every item in this guide exists to move that number.
The second outcome is recovery that is boring. Boring means the backups were verified last week. The restore procedure was rehearsed last quarter. The scope lists come from logs you kept, and nobody is improvising. A boring recovery still takes days, but they are predictable days with a known end. An exciting recovery is the one where you discover, on the worst morning of the year, that the offsite copy stopped syncing months ago. The difference is not budget. It is whether the controls are monitored or merely believed in.
The eight-signal detection set
Here is the full set, in roughly lifecycle order. Several are covered in depth elsewhere in this pillar or the library. This section is the index, with just enough on each to say what instrument feeds it and what the alert should be. Treat it as eight rows to fill in, not eight projects.
1. Beaconing
What: one internal host making small, regular connections to the same outside destination around the clock. Instrument: flow records or firewall connection logs, and DNS query logs, which show the same rhythm as repeated lookups of an unfamiliar name. Alert: a host whose connections to a single external destination show a steady interval over several hours, with near-zero payload, outside its baseline. The DNS side of this signal, including the odd-looking names some check-ins use, is covered in Beaconing and Domain-Generation Patterns in DNS.
2. Internal scanning
What: one host probing many internal addresses or ports in a short window — the discovery stage. Instrument: firewall deny logs between segments, flow records, and DNS logs showing one client resolving many internal names. Alert: distinct internal destinations per source per window far above that source's baseline, excluding your own scanner and monitoring server. The signatures and the allowlisting are in Internal Scanning: When One Host Starts Knocking on Every Door.
3. Lateral-movement fan-out
What: admin-protocol connections fanning out from a workstation or non-management server to many hosts, with matching authentication bursts on the domain controllers. The protocols are remote desktop, file sharing, and remote management. Instrument: flow records, inter-VLAN firewall logs, domain controller logon events. Alert: distinct admin-port destinations per source in ten minutes above five times baseline and at least ten, confirmed by a second window, a second signal, or the hour. This is the highest-value row in the set and has its own guide, Lateral-Movement Traffic: The Loudest Quiet Signal.
4. File-server activity storms
What: the encryption stage as seen by the file server: rename and extension storms from one account, symmetric read-then-write traffic from one client, and disk and interface saturation at an unscheduled hour. Instrument: file-server audit logs, flow records, SNMP or WMI server metrics. Alert: renames per account per minute above a hard floor; saturation with no scheduled job; both tightened outside business hours. Detail in Mass Encryption Signatures.
5. Backup health and immutability
What: backup jobs failing or missing, retention shortened, snapshots or shadow copies deleted, backup credentials used interactively or from the wrong host. Then comes the item most teams forget: the immutable or offline copy confirmed to still exist and still be unreachable from the production network. Instrument: backup system reports and logs, storage snapshot counts over SNMP or the storage system's own alerts, domain controller logon events. Alert: any job failure; any missing job report by the expected time; any retention or snapshot-count decrease outside a change window; any interactive use of the backup account. The verification half of this row gets its own section below.
6. Canary files
What: decoy files on each share that nothing legitimate ever touches. Instrument: per-file auditing on the decoys. Alert: any write, rename, or delete — critical, no confirmation, straight to a person. This is the cheapest row in the set and the only one with essentially no false positives. It fires late, once encryption has reached the decoy, but it fires with certainty.
7. Endpoint protection disabled
What: the antivirus or endpoint protection service on a machine stopping, being disabled, or having its real-time protection turned off — and the same for the host firewall. Attackers routinely switch protection off before the noisy stages, and the switching-off is itself logged. Instrument: the endpoint protection console, if you have a central one, and Windows service-state monitoring over WMI, which most infrastructure monitoring tools already do. Alert: protection service not running on any machine where it should be, with a short confirmation delay to ride out reboots. Protection disabled on several machines within an hour is treated as critical. One machine's protection going off is a support ticket; five machines in an hour is stage four of the lifecycle.
8. Monitoring agents going quiet in clusters
What: the monitoring system itself losing contact with several machines at once — agents stopping, WMI queries failing, SNMP timing out — while those machines still answer ping. During encryption, and sometimes in preparation for it, monitoring and management agents get stopped, and the monitor notices before anyone else can. Instrument: your monitoring system's own collection status. Alert: one agent going quiet is ordinary; three or more within a short window, on machines that still answer ping, is critical. Most monitoring tools can express "collector failed but host is up" as a distinct state; use it.
Notice which rows need no new instruments at all. Rows four, five, seven, and eight are largely expressible in ordinary infrastructure monitoring: server metrics, service states, backup reports, and collector status. Row six needs an afternoon. Rows one through three need flow data or log collection — the one meaningful investment in the set, and it pays off at every stage of the lifecycle.
Backup verification as a monitored control
A backup that has never been restored is a hypothesis. Most teams know this and intend to test restores regularly. Most do not, because nothing reminds them and nothing complains when they skip it. The fix is to make the restore test a monitored control. That is something the monitoring system expects to see on a schedule and alerts on when it does not. That is exactly like a nightly backup job.
The mechanism is the deadman check, also called a heartbeat. Instead of alerting when something bad is reported, you alert when something good fails to be reported. Each restore test, when completed, writes a timestamped result somewhere the monitor can read. That might be a log line, a file on a share, or a status update to the monitoring system. The monitor holds a rule: if no successful restore-test result has arrived within the expected interval, raise an alert. The test can be small as long as it exercises the real backup and the real restore path. For example, restore one folder from last night's backup to a scratch location and verify a checksum. Here is the shape of the rule:
Control: weekly restore test, FS01 Finance share -> scratch volume
Evidence: restore-test.log gains a line "OK <timestamp> checksum-match"
Rule: newest OK line older than 8 days -> WARNING (test is overdue)
newest OK line older than 15 days -> CRITICAL (control has lapsed)
newest line is FAIL -> CRITICAL (restore is broken)
Also: daily immutable-copy check:
expected retention count present AND repository NOT reachable
from a production-network test host -> OK
count decreased OR reachable from production -> CRITICAL
Three details make this work in practice. The test must be owned by a named person and scheduled on a calendar. That is because the monitor can only complain about a missed test, not perform it. The evidence must be written by the test itself, not by hand, or the control decays into someone typing "OK" on Friday afternoons. And the immutable-copy check must include the unreachability half. The point of an offline or immutable copy is that nothing on the production network can alter it. And the only way to be sure is to try from the production side and confirm failure. A backup your domain administrator account can delete is a backup the attacker can delete once they hold that account.
Once a quarter, scale the test up: restore a whole server or share to an isolated network and time it. The number you get is your honest recovery time for that system, and it belongs in the plan leadership sees. Vague recovery promises come from never having measured; boring recovery comes from having measured and written it down.
Around-the-clock routing for the top signals
Most of the alert-design advice in this library is about restraint. That means fewer alerts, higher thresholds, quiet hours, nobody paged for a disk that is merely getting full. That advice is correct, and ransomware signals are the exception. This is the one category where a missed alert can cost the organization its survival. It is where the attacker deliberately chooses the hours nobody is watching. And it is where the window between the first loud signal and irreversible damage can be minutes. If your monitoring wakes people up for anything, it should wake them up for this.
Practically, that means a separate routing path for a short list of critical signals. These are lateral-movement fan-out confirmed, canary touched, rename storm, snapshot deletion, protection disabled in a cluster, agents quiet in a cluster. The route bypasses quiet hours, bypasses maintenance-window suppression, and reaches a human through a channel that makes noise. Email is for morning reading. These alerts go by phone call or a push notification with its own sound. They go to an on-call person who has agreed to be on call and has the access to act. In a very small team the "rotation" may be two people alternating weeks, with a fallback if the first does not acknowledge within a few minutes. That is still a rotation, and it is far better than an alert that lands in an inbox at 2 AM and is read at 8.
Keep the list short and keep it honest. If the rename-storm rule fires three times a week on legitimate migrations, it will be muted within a month. Then the mute will still be there on the night that matters. Every rule on the wake-up list has to earn its place by staying quiet when nothing is wrong. That means tuning thresholds and maintaining allowlists as deliberately as firewall rules. A quarterly review of "what woke someone up, and was it real?" keeps the list trusted.
Tip: the person who gets woken up must be able to shut a switch port and disable an account from wherever they are. They must be able to do it within minutes, without a second person's password. Test that on a normal evening. An alert that reaches someone who cannot act is a notification, not a response.
The first-hour tabletop drill
A tabletop drill is a rehearsal with no real systems touched. The team sits down. Someone reads out a scenario one step at a time. Everyone says what they would do, who they would call, and where they would look. The drill takes an hour, costs nothing, and is the most effective readiness exercise for a small team. It turns "we have a checklist" into "we have done this." The checklist is The First Hour: Ransomware Response for Small Teams. The drill is walking through it against a made-up morning.
Run it like this. One person plays the narrator and has a scenario written in advance. A ransom note is found on the finance share at 6:50 AM, say. In that scenario, there is a lateral-movement alert from the night before that nobody acted on. The narrator reveals facts only when someone asks the right question, which is where the learning happens. Everyone else plays themselves. Walk the checklist in order, and at each step ask the questions that expose gaps:
- Isolate: which switch is that machine on? Who has the switch password? How long does it take to find the port? Can it be done from home?
- Protect backups: where is the offline or immutable copy? Who can confirm it is intact from an independent device? When was it last verified — and is that verification logged?
- Preserve evidence: how long do the firewall, DNS, and logon logs keep? Can they be exported somewhere off the production network, and does anyone know how?
- Communicate: who in leadership gets the first call, and is their number somewhere other than the file server? What does the user notice say?
- Engage help: is there an incident-response retainer or provider? Where is the insurance policy, and what is its notification window? Who is the legal contact?
- Scope: which logs answer "which hosts, which shares, which accounts," and can someone actually run those queries?
Write down every question that produced a shrug. That list is your readiness backlog, and it is usually more valuable than the drill itself. Fix the top three before the next drill. Run the drill twice a year. Once a year make one run a technical rehearsal. Actually shut a test machine's port. Actually restore a share to an isolated network. Actually export a day of logs. The first drill is always humbling. The third is where the checklist starts to feel like a routine rather than a hope.
The readiness checklist
Score yourself honestly. "Partly" is a legitimate answer, and it is more useful than a generous "yes."
| # | Readiness item | Evidence it is done | Reviewed |
|---|---|---|---|
| 1 | File-server metrics (disk queue, throughput, CPU, interface) monitored with tighter outside-hours thresholds | Alert rules exist and have fired in testing | Quarterly |
| 2 | Backup jobs monitored, with missing reports treated as failures | A deliberately skipped job produced an alert | Quarterly |
| 3 | Offline or immutable backup copy exists and is verified unreachable from production | Daily check logged; unreachability tested | Monthly |
| 4 | Restore test scheduled, automated evidence, deadman alert when overdue | Log shows successful tests at the expected interval | Weekly test, quarterly full restore |
| 5 | Canary files on every major share, per-file auditing, critical alert | Touching a canary paged someone in testing | Quarterly |
| 6 | Endpoint-protection and host-firewall service state monitored; cluster rule | Stopping the service on a test machine alerted | Quarterly |
| 7 | Monitoring collector status: "agent quiet but host up" alert with cluster rule | Stopping agents on three test machines alerted | Quarterly |
| 8 | File-server auditing on for writes, renames, deletes; rename-rate rule per account | Bulk rename in a test folder alerted | Quarterly |
| 9 | Flow records or inter-VLAN firewall logs collected and retained for at least a month | You can answer "who did this host talk to last Tuesday?" | Monthly |
| 10 | Lateral-movement fan-out rule with baseline, confirmation, and allowlist | A test fan-out from a workstation alerted; allowlist reviewed | Monthly |
| 11 | Internal-scan and beaconing rules | Test scan from a workstation alerted; beacon rule tuned against baseline | Quarterly |
| 12 | Wake-up routing for the critical list: bypasses quiet hours and maintenance suppression, reaches a rotation | A test alert at night reached the on-call by phone | Quarterly |
| 13 | On-call person can isolate a port and disable an account remotely, alone | Done on a test machine from home | Semi-annual |
| 14 | First-hour checklist printed and stored off the file server; leadership, insurer, legal, and provider contacts included | Paper copy located in under a minute | Semi-annual |
| 15 | Tabletop drill run; backlog of gaps written and top three fixed | Drill notes and backlog exist | Semi-annual |
Fifteen items is a lot for a small team, so take them in the order given. Items one through eight are ordinary monitoring and a few afternoons. Items nine through eleven are the flow-data investment. Items twelve through fifteen are organizational, and they decide whether all the rest matters at 2 AM on a Sunday. A team that has honestly completed the first eight and the last four is in far better shape than most, even before a single flow record is collected.
Where to go next
The honest goal of this program is modest and enormous at the same time. It is to shrink the window between the first foothold and a human knowing about it from days to minutes. And it is to make recovery, should it ever be needed, a rehearsed and boring procedure rather than the worst week of the year. None of it requires a security operations center. It requires only the signals in this pillar, turned into alerts that reach someone who can act. It also requires a team that has practiced the first hour before it happens. Start with the lifecycle in The Ransomware Lifecycle from the Network's Point of View. Build the detection that matters most from Lateral-Movement Traffic. Plant the tripwires from Mass Encryption Signatures. And drill The First Hour until it is routine. The whole set lives under Early Warning Signs of Ransomware on the Wire.
Put this into practice with SysPeep Monitor
More than half of this readiness checklist is ordinary infrastructure monitoring done carefully. That means file-server metrics with night-time thresholds, backup jobs whose silence counts as failure, service states over WMI, and collectors that notice when agents go quiet. SysPeep Monitor covers much of that ground from one Windows machine. It provides SNMP, WMI, and similar collectors. It provides thresholds, dependencies, and alert rules designed to keep noise down. And it provides email and other notifications for the short list that should wake someone up. The free trial takes minutes to install.
Download Free TrialLearn About SysPeep